[daip] Re: return of a CALIB failure (31DEC04 AIPS)

[Eric Greisen]

Thinking about it - I am suprised that the current code did not zero
divide.  With 1 baseline there can't really be much of an rms to be
thinking about.  In fact it really should be zero.  So it probably was
extremely small due to mathematical round-off issues and so the data
could look as if it were very far away or not depending on more
round-off issues.  I suspect that the code should refuse to do any
robust or rms things unless the number of baselines is large enough.

2 antennas = 1 baseline = perfect solution in principle
3 antennas = 2 baselines = perfect
4 antennas = 6 baselines = only 2x over
5 antennas = 10 baselines = 2.5x over   <--- maybe here we start

I will work on this - it should be straightforward.

If your MNJ has been fixed - it was broken at one point - then
it should update tomorrow.

Eric Greisen